AI Daily Brief: 30 March 2026

30 March 2026

This Monday morning brings a significant shake-up in the AI landscape: OpenAI's dramatic Sora shutdown, a landmark UK copyright ruling, and Anthropic marching towards one of the biggest IPOs in history. Meanwhile, 355 million users woke up to a dark DeepSeek screen overnight. Here is everything you need to know to start the week informed.

OpenAI Kills Sora: The $1bn Miscalculation That Exposed AI's Compute Problem

OpenAI's decision last week to shut down Sora, its AI video-generation app, just six months after public release sent shockwaves through the industry. The app had promised to democratise video creation and had even secured a reported $1 billion partnership with Disney. Yet it is now gone, and the Disney deal with it.

The real story, according to a Wall Street Journal investigation cited by TechCrunch, is strikingly mundane: Sora was a money pit nobody was using. User numbers peaked at around one million and then collapsed to fewer than 500,000. All the while, the app was burning approximately $1 million per day because video generation is extraordinarily expensive to run. Every user creating a short clip was consuming a finite supply of high-demand AI chips.

Compounding the problem was a competitive one. While OpenAI had a team dedicated to making Sora work, Anthropic's Claude Code was winning over the software engineers and enterprises that actually drive AI revenue. Sam Altman made the call: shut Sora, free up compute, and refocus.

Our take: This is a cautionary tale for any business evaluating AI product investments. Consumer-facing generative AI Tools are extraordinarily costly to run and face brutal retention challenges once the novelty fades. The companies winning in AI right now are those solving concrete business problems, not the ones chasing the most visually impressive demos. The Sora shutdown also illustrates that even well-funded AI labs face genuine resource constraints. Compute is the new oil, and OpenAI just chose to drill where the return is higher.

UK Copyright and AI: Government Reverses Course on Opt-Out Model

In one of the most significant UK regulatory developments of the year, the government has abandoned its preferred position on AI training and copyright. A broad copyright exception with an opt-out mechanism, which would have allowed AI companies to train on UK content by default unless rights holders actively objected, is no longer on the table. The government says it will gather further evidence before deciding on any legislative changes.

The reversal comes after the House of Lords Communications and Digital Committee published a report on 6 March calling generative AI a "clear and present danger" to the creative industries. The Committee recommended the government reject the opt-out model entirely and instead strengthen creator rights, calling for greater transparency on what AI models are trained on and better licensing frameworks.

The government's own March 2026 position paper also proposed removing copyright protection for wholly computer-generated works while confirming that works created with AI assistance would retain standard copyright. It also flagged plans to regulate digital replicas and explore labelling of AI-generated content.

Our take: This is significant for any UK business involved in content creation, media, or building AI systems. The opt-out model would have been a significant green light for AI developers to use UK content freely. Without it, training data licensing becomes a live commercial and legal question. UK businesses building AI products that incorporate third-party content should review their legal exposure now, not when legislation eventually lands.

Rishi Sunak and Goldman Sachs Tell UK Small Businesses: Adopt AI or Fall Behind

Goldman Sachs brought together 300 UK small business leaders at a Birmingham library last week for its "10,000 Small Businesses" AI for Growth summit. The headline message, delivered in part by Rishi Sunak in his new role as a Goldman Sachs adviser, was clear: ignore AI at your peril.

Sunak framed AI adoption not as a technology choice but as a competitive necessity. His rules for business leaders centred on getting hands-on with AI tools, identifying specific productivity bottlenecks, and acting quickly rather than waiting for the market to settle. Goldman's bankers outlined productivity data suggesting SMEs adopting AI are already pulling ahead of those that are not.

Our take: The fact that Goldman Sachs is running dedicated AI summits for SMEs in regional UK cities tells you everything about where we are in the adoption cycle. This is no longer a conversation for enterprise technology teams. It is mainstream business strategy. If you are a UK small or medium business and AI is still on the "to do later" list, the gap between you and your AI-enabled competitors is widening every quarter. The question has shifted from "should we use AI" to "which processes do we tackle first".

Apple to Open Siri to Claude, Gemini and Rival AI Assistants in iOS 27

Apple is planning a significant overhaul of Siri in iOS 27 that would allow third-party AI assistants including Anthropic's Claude and Google's Gemini to operate through Siri via a new "Extensions" framework. Users would be able to enable or disable individual AI services through a dedicated settings menu across iOS 27, iPadOS 27, and macOS 27.

The move extends Apple's existing ChatGPT integration to a broader competitive landscape. Interestingly, Bloomberg also reports that Apple is building its own AI chatbot codenamed "Campos" powered by Google's Gemini, suggesting even Apple's in-house AI efforts will lean on external models.

Our take: This is a landmark shift. Apple has historically been fiercely protective of the iPhone platform, so opening Siri to competitors signals that AI capability is now a table-stakes feature Apple cannot deliver alone. For UK businesses developing AI applications or services, this creates a significant new distribution opportunity. Reaching iPhone users through Siri integration will become a meaningful channel in 2027. Start planning now.

Anthropic Targets $60bn+ IPO in Q4 2026

Anthropic, the maker of the Claude AI assistant, is in early discussions with Goldman Sachs, JPMorgan and Morgan Stanley about going public as early as October 2026. Bankers expect the offering to raise more than $60 billion, which would make it the second-largest IPO in history, behind only SpaceX.

The timing is notable. Claude Code is currently dominating AI-assisted software development. Claude's Computer Use feature, which lets the AI control Mac keyboards and mice directly, was launched this week. Anthropic is also building out an enterprise client list that includes major banks and law firms. The IPO would crystallise a company that was valued at around $60 billion in private funding rounds earlier this year.

Our take: An Anthropic IPO at this scale would be a defining moment for the AI industry, bringing institutional investor scrutiny and public accountability to a company whose safety-first positioning has distinguished it from competitors. For UK businesses, the practical implication is that Anthropic is not going anywhere. Organisations building on Claude can expect continued investment in enterprise features and stability. It also sets the stage for an AI sector that is increasingly mature and publicly accountable, which matters for governance conversations in large organisations.

DeepSeek Outage: 355 Million Users Left Dark for Seven Hours

DeepSeek, the Chinese AI startup that shook the market with its R1 model in January 2025, suffered its biggest outage since launch overnight. The chatbot was down for more than seven hours, affecting an estimated 355 million users. Services were eventually restored following multiple system updates, though the cause has not been disclosed.

The disruption is the most visible sign yet of the infrastructure pressures facing AI providers operating at scale. DeepSeek's R1 model attracted enormous attention partly because it appeared to achieve frontier-level performance at a fraction of the compute cost. Whether that efficiency claim holds under 355 million concurrent users is increasingly open to question.

Our take: For UK businesses that have incorporated AI tools into core workflows, this is a timely reminder to assess your resilience posture. If a single AI provider going down for seven hours would halt critical operations, that is a risk worth addressing. This is particularly relevant for firms relying on lower-cost providers to achieve AI efficiency gains: price and reliability are not always correlated. Build contingency into your AI architecture the same way you would any critical supplier relationship.

AI Has Broken the Software Org Chart: Product Managers Are Now Shipping Code

A first-hand account in VentureBeat describes the practical reality of running an AI-first software team in 2026: product managers are building and shipping features directly, designers are fixing their own CSS without filing a ticket, and engineers have shifted from writing code to validating and directing AI-generated output.

The author notes that once AI coding agents reduced Implementation cost to near zero, the bottleneck moved from engineering capacity to decision velocity. All the coordination mechanisms built to protect engineering time, such as specs, tickets, handoffs, and backlog grooming, became the drag on the system.

Our take: This is one of the most practically important AI-in-business stories of the moment, and it applies well beyond software companies. When AI agents can execute, the highest-value human skill becomes the quality of your thinking and the speed at which you make decisions. For UK organisations going through AI transformation, this should prompt a genuine review of internal processes: not just which tasks AI can assist with, but which coordination structures were built to compensate for human execution limits and can now be retired.

Quick Hits

• Bandcamp bans AI content: Bandcamp has become the first major music platform to prohibit AI-generated music, drawing a clear line for the creative industries.
• UK AI in energy: Centrica's Gunjan Arora argues in BusinessGreen that AI-driven grid optimisation is one of the most undervalued levers in the UK's net zero transition.
• UK AI lead under pressure: The BBC reports that the UK leads Europe in AI companies per capita, but warns of a brain drain to the US as entrepreneurs seek larger pools of capital.
• Legora raises $550m: Legal AI startup Legora has secured $550 million to accelerate US expansion, reinforcing the strong investment thesis for vertical AI in professional services.
• ARC-AGI-3 humbles every AI model: The new ARC-AGI-3 benchmark scored every frontier model below 1% on tasks that 100% of humans solved on their first attempt. Gemini 3.1 Pro led at 0.37%. The gap between AI capability and human general reasoning remains vast.
• Waymo hits 500,000 paid rides per week: Waymo has doubled its ride volume to 500,000 paid trips per week in under a year, already halfway to its end-of-year target.

Frequently Asked Questions

Why did OpenAI shut down Sora so quickly?

OpenAI shut down Sora six months after launch because user numbers had collapsed from around one million to fewer than 500,000, while the app was costing approximately $1 million per day to run. With compute resources under pressure and competitors such as Anthropic winning enterprise customers, CEO Sam Altman chose to redirect resources to higher-return products.

What does the UK government's position on AI and copyright mean for businesses?

The UK government has stepped back from plans to introduce a broad copyright opt-out for AI training data. This means AI companies cannot assume they can freely train on UK content by default. For businesses building AI systems that use third-party data, this increases the importance of reviewing licensing arrangements and data provenance. The situation remains in flux as the government gathers further evidence before any legislative changes.

How should UK SMEs approach AI adoption in 2026?

Start now, start small, and focus on specific productivity bottlenecks rather than broad transformation projects. Identify two or three repetitive, time-consuming processes in your business and pilot AI tools against those. Measure the time saved, then expand. Waiting for AI tools to mature further means falling behind competitors who are already building proficiency and efficiency gains.